Implementation:

        Break up the speech signal into windows of optimum length (say 30ms for eg)
         The length of the window depends upon the % by which the speech signal is to be decreased.
Shift each window by the required amount or by the nearest optimum amount of samples
            ie. If  we require to decrease the speed of the speech signal by 50% then  shif t the window by 50% of the
                window length.
        Find Optimum Shift
                  Vary the required shift amount from (Reqd shift - A) to (Reqd shift + A) and find the Meean Square Error. Use
                    that as the required shift.
        Add the Shifted window on to the original window to form a new window.
 
        Repeat the same with other windows.
        Add together all the new windows to form the new speech signal.